Exogenous Data in Forecasting: FARM -- An Approach for Relevance Evaluation. (arXiv:2304.11028v1 [eess.SP])

Exogenous data is believed to play a key role for increasing forecasting
accuracy. For an appropriate selection, a throughout relevance analysis is a
fundamental first step, starting from the exogenous data similarity with the
reference time series. Inspired by existing metrics for time series similarity,
we introduce a new approach named FARM - Forward Angular Relevance Measure,
able to effectively deal with real-time data streams. Our forward method relies
on an angular feature that compares changes in subsequent data points …

accuracy analysis angular arxiv data data streams deal evaluation exogenous feature forecasting metrics real-time reference role series time series